Prevalence of Donor-Specific Antibodies After Pediatric Liver Transplantation: A Meta-Analysis.

Our recent study showed that donor-specific HLA antibodies (DSAs) were common after pediatric liver transplantation (LT).[1] The aim was to conduct a meta-analysis of DSA prevalence after pediatric LT.

PubMed was used (as of March 1, 2016) with a search strategy: Human leucocyte antigen antibod* OR HLA antibod* OR donor-specific antibod* OR donor specific antibod* OR DSAs AND liver transplantation AND (pediatric OR children). Studies published before January 1, 2000, were excluded based on a rationale for HLA antibody detection technology advancement.

References were screened, and data were extracted from the eligible studies based on 3 criteria: (1) patients underwent LT <18 years of age, (2) prevalence of DSAs was evaluated after LT, and (3) patients were on some form of immunosuppression at the time of DSAs. Studies with mixed population of adults and children were included if information was available for pediatric patients separately or if average age of patients were under 18 years at LT. Unit of analysis was the proportion of patients with DSAs of total number of patients analyzed for DSAs.

R version 3.1.1 (www.r-project.org) was used with meta package[2] to calculate overall prevalence with 95% confidence interval (CI). The overall prevalence was calculated with the use of logit transformation. A random-effects model with the method by DerSimonian and Laird was used. The average prevalence estimate across the studies is obtained under the random-effects model. In addition, 95% prediction interval (PI) was calculated to obtain a predicted range of true DSA prevalence for a new analogous study. The details of PI have been described elsewhere.[3]

Literature search yielded 28 references of which 8 were included in the analysis. One study missed in the aforementioned search was included based on an earlier knowledge of its existence (total of 9 studies). Reasons for exclusion were mixed population (n = 5), intestinal transplantation (n = 5), case reports (n = 3), review (n = 2), editorial (n = 1), DSAs evaluated only prior LT (n = 1), overlapping study population (n = 1), patients not on immunosuppression (n = 1), and full text not available (n = 1). Some of the studies had multiple reasons for exclusion.

Age at the time of LT varied between studies (Table 1). Two of the studies included patients not on immunosuppression at the time of DSAs (Table 1 footnote).

TABLE 1

List of included studies

Total sample size was 322 patients (Figure 1). Average prevalence of DSAs was 41% (95% CI, 29%-54%) although prevalence varied across studies as evident with the heterogeneity statistics (I2 = 77% [95% CI, 56%-88%], P < 0.001 for heterogeneity). The PI indicated that the true DSA prevalence for a new analogous study will fall 95% of times within interval of 11% to 80%.

FIGURE 1

Average (95% CI) prevalence of 41% (29% to 54%) (grey diamond) for DSAs after pediatric LT. I2 is shown with 95% CI. I2 depicts that 77% of variability in DSA prevalence estimates across studies is beyond sampling error (ie, due to heterogeneity). CIs for individual studies were calculated with the method by Clopper-Pearson. Between-study variance (τ2) was estimated with the method by DerSimonian and Laird. The prediction interval (black solid line) refers to a predicted interval of true DSA prevalence for a new analogous study.

There was little influence on heterogeneity when omitting 1 study at the time. The largest impact on heterogeneity was observed after excluding study by Waki et al[4] (I2 from 77% to 66%) or study by Markiewicz-Kijewska et al[5] (I2 from 77% to 71%). Average DSA prevalence was 51% (95% CI, 43%-58%) after excluding these 2 studies simultaneously. Heterogeneity also diminished (I2 = 29% [95% CI, 0%-69%], P = 0.207), and the PI also became narrower as expected (34% to 67%). The other excluded study evaluated DSAs shortly after LT (at 3 weeks).[4]

This study has limitations. First, only 1 author extracted the data which can bias the extraction process. Second, the literature search was simple and based only on 1 database. Third, the impact of different study characteristics on heterogeneity was not assessed.

These limitations in mind, the average DSA prevalence was 41% (95% CI, 29%-54%) across 9 studies after pediatric LT although variability between studies was noticeable.